Hi!

In PR119491 r15-9154 I've allowed some useless EH regions for musttail
calls (if there are no non-debug/clobber stmts before resx which resumes
external throwing).
Now, for -O1+ (but not -O0/-Og) there is a cleanup_eh pass after it
which should optimize that way.
The following testcase ICEs at -O0 though, the cleanup_eh in that case
is before the musttail pass and dunno why it didn't actually optimize
it away.

The following patch catches that during expansion and just removes the note,
which causes EH cleanups to do the rest.  A tail call, even when it throws,
will not throw while the musttail caller's frame is still on the stack,
will throw after that and so REG_EH_REGION for it is irrelevant (like it
would be never set before the r15-9154 changes).

Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

2025-04-04  Jakub Jelinek  <ja...@redhat.com>

        PR middle-end/119613
        * cfgrtl.cc (purge_dead_edges): Remove REG_EH_REGION notes from
        tail calls.

        * g++.dg/opt/pr119613.C: New test.

--- gcc/cfgrtl.cc.jj    2025-02-13 19:59:55.165577958 +0100
+++ gcc/cfgrtl.cc       2025-04-04 09:17:16.309666732 +0200
@@ -3213,6 +3213,16 @@ purge_dead_edges (basic_block bb)
              && ! may_trap_p (XEXP (eqnote, 0))))
        remove_note (insn, note);
     }
+  /* A tail call cannot trap either.  The tailc/musttail pass could have
+     allowed a tail call if it could throw internally, but perform no
+     actual statements and then caused the exception to be thrown externally
+     in the hope that it is cleaned up later.  If it is not, just
+     remove REG_EH_REGION note.  While the call maybe can throw, the
+     current function's frame will not be there anymore when it does.  */
+   if (CALL_P (insn)
+       && SIBLING_CALL_P (insn)
+       && (note = find_reg_note (insn, REG_EH_REGION, NULL)))
+     remove_note (insn, note);
 
   /* Cleanup abnormal edges caused by exceptions or non-local gotos.  */
   for (ei = ei_start (bb->succs); (e = ei_safe_edge (ei)); )
--- gcc/testsuite/g++.dg/opt/pr119613.C.jj      2025-04-04 09:16:56.867680864 
+0200
+++ gcc/testsuite/g++.dg/opt/pr119613.C 2025-04-04 09:18:24.541617133 +0200
@@ -0,0 +1,22 @@
+// PR middle-end/119613
+// { dg-do compile { target { musttail && { c || c++11 } } } }
+// { dg-options "-O0" }
+
+struct S { S () {} };
+char *foo (S);
+void bar (int);
+
+[[gnu::always_inline]] inline char *
+baz (S x)
+{
+  unsigned i;
+  &i;
+  bar (i);
+  [[gnu::musttail]] return foo (x);
+}
+
+char *
+qux (S)
+{
+  [[gnu::musttail]] return baz (S {});
+}

        Jakub

Reply via email to